The wonderful wizard of LoC: paying attention to the man behind the curtain of lines-of-code metrics

Abstract

Lines-of-code metrics (loc) are commonly reported in Programming Languages (PL), Software Engineering (SE), and Systems papers. This convention has several different, often contradictory, goals, including demonstrating the 'hardness' of a problem, and demonstrating the 'easiness' of a problem. In many cases, the reporting of loc metrics is done not with a clearly communicated intention, but instead in an automatic, checkbox-ticking, manner.In this paper we investigate the uses of code metrics in PL, SE, and System papers. We consider the different goals that reporting metrics aims to achieve, several various domains wherein metrics are relevant, and various alternative metrics and their pros and cons for the different goals and domains. We argue that communicating claims about research software is usually best achieved not by reporting quantitative metrics, but by reporting the qualitative experience of researchers, and propose guidelines for the cases when quantitative metrics are appropriate.We end with a case study of the one area in which lines of code are not the default measurement-code produced by papers' solutions-and identify how measurements offered are used to support an explicit claim about the algorithm. Inspired by this positive example, we call for other cogent measures to be developed to support other claims authors wish to make. CCS Concepts: • General and reference → Metrics.

Employee Empowerment and HR Flexibility in Information Technology SMEs

“…37,38 Further, narrow performance metrics used for performance management, such as numbers of lines of code or agile project velocity, require considerable re-configuration effort and time when software languages and development frameworks change or teams are re-configured. 39,40 Similarly, an excessive emphasis on narrowly-defined IT certifications incurs maintenance costs when technology changes or the firm switches to using a different technology. 41 At the same time, generic HR practices such as management by objectives 42 or socialization (social activities aimed to facilitate knowledge sharing), 43 are applicable in a very broad range of situations, and thus are robust in the face of change and can be easily configured to promote or facilitate change (e.g., by defining objectives in terms of change or by arranging events to facilitate tacit knowledge exchange between groups of employees who need to collaborate to effect change).…”

Code Red: The Business Impact of Code Quality -- A Quantitative Study of 39 Proprietary Production Codebases

“…There are numerous studies on code quality metrics [5,52]. Although primarily used to communicate how comprehensive a software solution is or how much work a development task would require, Lines of Code (LoC) metrics are often involved also in code quality metrics [2]. An underlying assumption is that the higher LoC a file has file, the more complex it is.…”

“…The Code Health metric is modelled around the idea that the most promising approach to measuring code complexity is by identifying and measuring specific attributes of complexity [22]. As such, Code Health incorporates elements from the generic ISO 5055 maintainability section, and complements them with design-level code smells 2 . Examples of included design smells are God Class, God Methods, and Duplicated Code [34].…”

Could Tierless Languages Reduce IoT Development Grief?

“…(3) To ensure correctness the developer must maintain type safety across a range of very diferent languages and diverse type systems. (4) The developer must deal with the potentially diverse failure modes of each component, and of component interoperations.…”

“…We systematically compare code size (source lines of code (SLOC)) of the four smart campus implementations as a measure of development efort and maintainability [4,63]. The tierless implementations require 70% less code than the tiered implementations.…”